Skip to main content
Minovoice · local transcription

Speech to text that gets Korean
right.

Meetings, interviews, phone recordings — transcribed on this Mac, never in the cloud. Speakers separated, every sentence editable next to its audio, exported straight to subtitles or minutes. Powered by BatiSay: Korean speech models built in-house, not another Whisper wrapper.

macOS 14+ · Apple Silicon100% on-deviceUnlimited files · unlimited length
Minovoice
team-meeting_0711.m4a · 1:24:03
화자 2명 감지 · 전화 모드 꺼짐100% on-device
화자 1

그럼 3분기 출시 일정부터 정리할게요.

00:41
화자 2

네, 자막 팀 쪽은 SRT 내보내기가 제일 급해요.

00:47
화자 1

COSRX 건은 이번 주에 계약서 검토가 끝납니다.

00:55
화자 2

통화 녹음 파일도 같은 폴더에 넣어둘게요.

01:02
SRTVTTMDDOCXTXT
폴더 감시 중 · 3 files queued

Most Mac transcription apps share the same engine

It reads English fine. Conversational Korean is where it falls apart.

Whisper-based apps stumble on real Korean conversation — meetings, calls, interviews.

Minovoice runs BatiSay, Korean speech models built in-house for exactly that audio.

Cloud transcription caps your minutes and takes your audio with it.

No monthly limits, no upload. Audio and text never leave this Mac — decisive for legal, medical, HR work.

A 2-hour recording dies at minute 40.

Long files are the normal case here — 2–3 hour recordings transcribe stably, end to end.

The transcript lands in your editor as a wall of unattributed text.

Speakers come pre-separated, sentences play their audio on click, and fixes happen inline.

From recording to finished document

Transcription is step one. Minovoice covers the rest of the job.

In-house model

Powered by BatiSay

Korean speech models trained in-house by Bati AI — the part of the product nobody can copy by wrapping the same open-source engine. Not another Whisper wrapper.

100% on this Mac

Audio in, text out — all inference local. Nothing uploads, which means recordings that must not leave the building, don’t.

No meters running

No monthly minutes, no file counts, no length caps. Batch a folder of recordings and walk away.

Speakers, separated

Who said what, color-coded and renameable. A dedicated phone-recording mode handles 8kHz two-person call audio.

An editor that finishes the job

Click a sentence to hear that exact moment. Fix inline. Teach it your proper nouns — right-click a mistake once and it is corrected everywhere, forever.

Five exports + folder watch

SRT and VTT subtitles, Markdown, DOCX, plain text, clipboard. Point it at a folder and new recordings transcribe themselves.

How Minovoice compares

Local apps miss Korean. Korean apps live in the cloud. Pick both instead.

MinovoiceCLOVA NoteSuperwhisperMacWhisper
Conversational Koreanown model
Local / privatecloud
Unlimited minutes300 min/mo
Speaker separation+ phone modepartial
Built-in correction editorbasicbasic
Price$49 early birdsubscription$249$79

Sold and supported by minodeck. Speech models by Bati AI.

One license. No meter.

The trial is the full product on 3 files — quality is the pitch, so you get all of it.

Trial

$0
3 files · every feature, full quality
  • Full BatiSay Korean models
  • Speaker separation + phone mode
  • Complete editor + dictionary
  • All five export formats
  • No watermarks, no quality caps
Try 3 files free

Full license

Early bird
$49
one-time early bird · $69 after — v1 lifetime license (major upgrades separate)
  • Unlimited files, unlimited length
  • Speaker separation + phone-call mode
  • Editor with click-to-play + inline fixes
  • Personal pronunciation dictionary
  • SRT · VTT · Markdown · DOCX · TXT + folder watch
  • All v1 updates included
Get Minovoice

Roughly one year of a transcription subscription — except it does not expire.

Frequently asked questions

No — v1 transcribes recorded files (audio and video). Live dictation is a different product with different tradeoffs, and we would rather do file transcription excellently first.

Hear it in text.

Three files free — full quality, every feature. Judge the Korean yourself.

100% on-device · No subscription · v1 lifetime license